13th World Conference on Earthquake Engineering

Vancouver, B.C., Canada

August 1-6, 2004
Paper No. 645

RELIABILITY BASED ANALYSIS AND DESIGN OF ANCHOR

RETROFITTED CONCRETE GRAVITY DAMS
Masoud R. KAZEMI1

SUMMARY
This paper presents a comparison between level of safety provided by traditional allowable stress
evaluation of monolithically built concrete gravity dams and a more recent Ultimate Limit States or
Reliability Based evaluation of these gravity based structures subject to seismic and non-seismic loads.
Level of safety resulting from a conventional seismic stability analysis of a typical concrete gravity block
subject to normal static loads and pseudo static earthquake load is compared with the level of safety
determined from reliability or ultimate limit states design principles. The tallest monolith of the Pine Flat
Dam is used as an example case for calculation of traditional sliding factor of safety and reliability indices.
The dam monolith with and without post tensioned anchors are considered. This paper demonstrates that
while some measure of safety is ensured by following the conventional factor of safety approach in the
stability analysis and subsequent remedial design, reliability based analysis and design of stabilized
concrete gravity dams provides a consistent level of structural reliability in the stability analysis of such
structures, in particular where remedial design is required.
INTRODUCTION
Recent publication of dam safety guidelines and research around the world reflect a growing concern on
the seismic safety of hydroelectric and flood control gravity dams and other similar water retaining gravity
structures built prior to 1950s. Although probable maximum flood (PMF) is a critical lateral load in the
sliding stability of concrete gravity dams, by and large horizontal and vertical accelerations due to
maximum credible earthquake (MCE) have become the main factor in the stability evaluation of concrete
gravity dam structures.
A typical medium to large size concrete gravity dam comprises various service or relief structures that
include a number of expansion joints. By controlling temperature ingredients, expansion joints are mainly
provided to minimize cracking of concrete mass due to heat of hydration of concrete during and
immediately after construction of dams. Inclusion of expansion joints results in the creation of several
rigid concrete monoliths or blocks along the length of the dam. During a strong seismic event in the
upstream-downstream direction (i.e. along the centerline of river), concrete blocks tend to slide at the base
1

President, Hannah Kazemi + Company LTD., Vancouver, BC, Canada, Email: mkazemi@hannahcanada.com

of concrete block or rock independently between the joints, thus resulting in a potential breach in the dam
structure. A potential dam breach would result in endangering the public safety and also immense
economic loss. To remedy this problem, post tensioned multi strand rock anchors are typically employed
to increase the stability of portions or all of a hydroelectric or flood control dam structure. This paper
presents reliability based principles applied in assessing the safety of gravity dams and design of post
tensioned anchors where dam stabilization is required.
MECHANICS AND EFFECTS OF POST TENSIONED ANCHORS
In most dam rehabilitation projects, anchor installation to restore dam stability is deemed the least cost
and most time effective option when compared against other options such as adding substantial volume of
concrete to main body of dam or constructing alternative water passages to alleviate water pressure on the
dam. A rock anchor is comprised of a group of high strength strands that are bundled together and
designed to the required post tensioning or stabilizing force. Since multi strand anchors are the primary
component in dam stabilization, a typical anchor is examined for its mechanical properties and behavior.
In general, the beneficial effects of rock anchors can be examined by considering the installation process
and composition of rock anchors, which are as follows:
1) Drill anchor hole to the designed diameter and depth;
2) Water pressure test the drilled hole by filling up the hole to the top and observing whether water
level lowers down or not. A drop in water level indicates the presence of fissures and cracks in
foundation rock material, which as a remedy, drilled hole is grouted and later drilled to the
designed diameter and depth. The water pressure test is again conducted and if the test fails
again, the grout, drill and water pressure test process is repeated until water pressure test passes;
3) Place shop fabricated anchor in drilled hole. Place and fill stage 1 grout to the designed bond
length which is within the rock foundation;
4) After stage 1 grout has reached the designed compressive strength, post tension the anchor to the
designed stress level (e.g. 0.80fsu) and lock off at anchor head.
5) Place and fill stage 2 grout by completely filling the anchor hole with grout and cap off the anchor
head.
It should be noted that the main purpose of the grouting operation is to fill the drilled hole, without any
voids or cavities, with a cement based material to act as a corrosion protection layer for the post tensioned
steel strands. The grout material should also possess sufficient bond strength to transfer strand skin or
bond stresses to the surrounding concrete or rock material. Cost permitting, strain gauges could be
installed along the length of anchors to allow for short and long term monitoring of bonded post-tensioned
anchors. The presence of the designed post tensioning force or any reduction in force due to possible long
term relaxation in anchors can be verified by strain gauges.
In the past, structural engineers relied upon factors of safety in the design and analysis of structures, in
particular hydroelectric and gravity dam structures. Although the use of common safety factors provided
some assurance of safety in a particular design situation, it did not, among other factors, provide a
consistent level of structural reliability in the design. The basic concept is straightforward: factor of safety
of some quantity is simply the ratio of the allowable value, called Capacity (C), to the calculated value,
called Demand (D), and is generally expressed as:

[1]

FOS =

C
D

The widely used Working Stress Design method (WSD) uses allowable and calculated stresses from
applied loads in a particular structural member or component, and ensures that equation [1] is greater than
a predetermined value of the safety factor, FOS. In the past, and to some extent even at present, subjective
judgment by engineers was used to select a proper factor of safety in design. Invariably, this approach did
not explicitly take into consideration the variability in loads and material strength.
Principles of probability theory, in the form of reliability analysis have recently been utilized to minimize
the high variability in safety that is inherent in the simple Working Stress Design method. The concept of
reliability based design has now been adopted by most industrialized countries in the form of Limit States
Design (LSD) or Load and Resistant Factor Design (LRFD). Ongoing research is being conducted to
determine accurate load and resistance factors. Recent literature ( Hasofer, Lind, 1974; Allen, 1975;
Foschi, 1978; Sexsmith, Fox, 1978; Sexsmith, 1979; Whitman, 1984; Foschi, Folz, Yao, 1989 and 1993;
etc.), is briefly reviewed here.
In general, a structural design problem involves the interaction of several random variables which mainly
includes the geometric and material properties of the structure, and can be mathematically formulated as a
vector of basic random variables, in the form of:

[2]

X = { X 1 , X 2 , X 3 , ..., X n }

The limiting state and performance of the structure, in terms of X , may then be described by a
performance function in the form of:

[3]

G{ X } = { X 1 , X 2 , X 3 , ..., X n }

To differentiate between Capacity and Demand random variables, as it is the usual case in structural
design, equation [3] can be re-written as:

[4]

G =C D

Therefore, failure of structure is imminent when G < 0 (i.e. demand is greater than capacity), and
conversely, survival of the structure is indicated by G > 0 . Then, the boundary between failure and
survival, G = 0 , can be considered as a limit state surface between all the variables involved in the
performance function [3]. From a structural design point of view, a limit state is defined by various states
of collapse and un-serviceability that are to be avoided (Allen, 1974). Ultimate limit states relate to the
safety and integrity of the structure, and relate to the load carrying capacity of the entire structure or part of
it. Material fracture, structural instability and excessive crushing are some typical cases that are
encountered in structural design. On the other hand, serviceability limit states correspond to those cases
that relate to the level of comfort in a structure under specified loads. Common examples are floor
vibration, excessive floor deflection and cracking. The work presented herein relate to sliding movement
of concrete gravity blocks which ultimately could lead to dam breach and as such governed by ultimate
limit states principles.
In general, an entity's reliability, by definition, is the probability that it will remain fully functional
throughout its design life span. For a structural entity, reliability can be viewed as the complement of the
probability of failure, Pf , which is given by the multiple integral:

[5]

Pf =

1, 2 , 3,...,n

( X 1 , X 2 , X 3 , ..., X n )dX 1 dX 2 dX 3 ... dX n

Failure

where the failure domain, Failure , is expressed as G{X} 0 , and the integrand is the joint probability
density function of the intervening variables. Alternatively, equation [5] may be simply expressed in
terms of the vector of basic random variables X :

[6]

Pf =

X
G{ X }< 0

( X )dX

On the other hand, the reliability or probability of survival is the volume integral over the safe region:

[7]

Pf =

X
G{ X }> 0

( X )dX

Since a high number of random variables define a typical structural problem, the joint probability density
function of all involved variables is difficult to model, and if indeed the function exists, then the closed
form solution of the governing integral [6] or [7] becomes very complicated. Aside from approximate
Monte Carlo simulation techniques, a commonly used approach to this reliability evaluation procedure is
to work with the so called reliability index, , which is based on approximate FORM (First Order
Reliability Method) and SORM (Second Order Reliability Method) procedures. Thus, the probability of
failure, Pf , may be estimated by the standard normal (or Gaussian) probability distribution function,
(.) :

[8]

Pf = ( )

It should be noted that if all participating variables are normally distributed with known model parameters
and the performance function [3] is linear, then the probability of failure [8] can be determined exactly.
However, to generalize the analysis, by knowing the mean, X i , and standard deviation, X i , of each
involved variable in the equation [3], and introducing reduced variate i = X i X i X i , for
i = 1, 2 , 3,..., n , the reliability index, , as depicted in Figure 1 for the simple case of two variables, can
be established as the minimum distance from the origin to the limit states surface. The intersection point
between this minimum distance and the limit states surface may be called the design point, * , and can be
found through FORM/SORM techniques (Foschi et al. 1989).

Figure 1 - Mathematical interpretation of Reliability Index.

As shown in Figure 1, FORM approximates the limit condition by fitting a tangent plane to the limit states
surface at the design point, while SORM utilizes a quadratic surface to better approximate the true limit
state. In addition, to refine the approximation of the probability of failure, Pf , it is required that the
related variables be non-correlated. Since most structural problems involve correlated variables, a proper
transformation is required to acquire uncorrelated variables and consequently estimate Pf (Der
Kiureghian and Liu 1986).
A computer program RELAN (RELiability ANalysis) was developed by Foschi, Folz, Yao (1989) to carry
out the above analysis and calculations. RELAN includes the Rackwitz-Fisseler algorithm for non-normal
variables and transformation of correlated variables by Der Kiureghian and Liu (1986). Since the work of
this paper is focused on the reliability aspects of concrete gravity dams, this program is considered to be of
use to compute and establish reliability indices for not reinforced and anchor reinforced gravity dam
monolith under various load cases.
The level of safety comparison between not reinforced and anchor reinforced cases is best illustrated by a
case example shown herein.
GRAVITY DAM WITHOUT ANCHOR REINFORCEMENT
While concrete gravity dams are generally designed and built to suit specific geological and hydrological
site conditions, or for instance, in the case of power generation, to suit specific electric power demand, the
exact concrete outline and dimensions of concrete gravity dams around the world vary for each individual
dam. Items such as concrete and foundation rock structural properties and weight of associated structures
(bridges, gates etc.) are site specific and are required for the stability calculations. However, when
conducting a static or pseudo static stability analysis, stabilizing and destabilizing loads and forces acting
on gravity dams are generally similar in nature and, for illustration, only a typical non over-flow concrete
gravity section is considered herein. There is ample literature on the stability analyses of not reinforced
gravity dams and as such stabilizing and destabilizing loads acting on a typical gravity block are briefly
reviewed herein. The rigid gravity section shown in Figure 2 is primarily subjected to the following
destabilizing loads:
1. Hydrostatic pressure of upstream reservoir water and downstream tail water.
2. Silt pressure. In static conditions, submerged weight of silt is considered, while in dynamic
conditions, silt is considered to be liquefied and thus would exert an equivalent liquid
pressure.
3. Ice load, if any, which acts as a uniform distributed load at the upstream reservoir surface.

4. Added mass of water due to hydrodynamic pressure of upstream reservoir water.

5. Vertical uplift and horizontal earthquake inertial loads generated within the body of dam and
associated attachments, such as bridges, gates, hoists and cranes, if any.
6. Vertical uplift pressure at the base of the concrete gravity monolith. The magnitude of the
uplift pressure depends on the presence of drain holes at the base of the dam. Because drain
holes get plugged with sediment migration over time, depending on specific circumstances,
some degree of drain effectiveness is usually assumed in stability analysis.
The following loads are stabilizing loads considered in the stability analysis:
1. Self weight of concrete gravity block and associated attachments such as bridges, gates, hoists
and cranes, if any.
2. Downward vertical component of hydrostatic pressure of water or silt if the upstream face of
dam is inclined.
3. Downward vertical component of seismic inertial loads, if applicable in load combination
under consideration.

Y
Headwater

Ice

1
7
massai= w [H1 (H1 zi ] 2 . Ai
8

H1

mav +

mah

C.G.

mah +
mg ; mav

hs

7
H . .A
8 1 w i

Silt
(liquefy
under
e/q)

Tailwat er
C
L Drains

Ks s hs

H2

w H2

w H2
w H1

( H1 H2 )

w H2 +

Figure 2 Typical non-overflow concrete gravity dam monolith

It should be noted that, depending on load combination under consideration, the most unfavorable and
possible combination of above loads should be considered in the stability analysis of concrete gravity

dams. The beneficial effect of downward seismic inertial component, for instance, may be neglected as
the opposite upward component has a destabilizing effect, which usually is considered in the critical load
combination.
EXAMPLE PROBLEM
The application of reliability analysis is best illustrated by considering a typical gravity dam subject to
primary loads. The tallest monolith of the 561 m long Pine Flat Dam is considered. A simplified two
dimensional finite element model of this monolith is shown in Figure 3. A conventional sliding stability
analysis of the same monolith was conducted for illustration purposes. It should be noted that there are
relatively ample studies of the Pine Flat dam in the available literature and thus considered for this study.
C
LAnchor
Reservoir

Panchor
10.745 m

m
9
2
1
.
6
1
1
=
H

m
5
3
.9
4
1

0.78

0.05
1

Added Mass
of Water

mav

mah

C.G.

mah

m
5
8
9
.6
0
1

mg ; mav

Typical Nonlinear
Friction Element

95.805 m

Figure 3 Model of Pine Flat Dam

The commercial finite element package SAP2000 was used to conduct a non-linear time history analysis
to obtain maximum dam crest deformations, base sliding and shear time histories due to a horizontal
component of the Taft Earthquake, obtained at the Lincoln School Tunnel in the 1952 Kern County
Earthquake. The earthquake record was scaled up to a peak horizontal acceleration of 0.4g. The effective
uplift pressure is considered to be 40% of the total uplift due

The tallest monolith is idealized by shell elements with the base horizontal. The dam concrete is assumed
to be isotropic, homogeneous and elastic with a constant modal damping of 0.10. A series of biaxial
friction link elements were used to model the gap and friction behaviour between rock and concrete
surfaces. The reservoir bottom and foundation rock were assumed to be rigid. Hydrodynamic forces due
to earthquake were considered by considering the added mass of water (Westergaard) calculated at point i
along the upstream face of the dam, as follows:.
1
7
m ai = w [H ( H z i )] 2 Ai
8

[9]

where H = depth of water, z i = height above the base of the dam and Ai = tributary surface area at point
i. Because of reservoir height of about 116.3 m, fluid compressibility was taken into consideration by
increasing the above added mass values by 30%.
Analysis ResultsSliding Factor of Safety
A traditional factor of safety analysis of the tallest monolith of the Pine Flat Dam is conducted and
summary of forces acting on the horizontal sliding plane are shown in Table 1.
Table 1 Summary of forces to estimate sliding factor of safety
Force Component
Concrete

vert (kN)
42,153

Water on upstream face

Vertical uplift

horiz (kN)

987
-6,653

Hydrostatic thrust

20,162

Hydrodynamic thrust

8,490

Horizontal earthquake

16,861

Totals

36,487

45,513

The factor of safety against sliding is thus calculated as:

[10]

Factorof Safety against Sliding =

tan( ) Vert

Horiz

tan(45) 36,487
= 0.80
45,513

Based on above, stabilizing measures are required. This can be provided either by installation of vertical
or inclined post tensioned rock anchors, change in concrete profile by increasing its area or improvement
in foundation drainage to relive foundation uplift pressure. To meet a minimum acceptable factor of
safety against sliding, a stabilizing force of 9000 kN is required.
Analysis ResultsFinite Element Analysis

A nonlinear time history modal analysis of the tallest monolith of the Pine Flat Dam subject to the scaled
Taft earthquake record reveals that significant upstream downstream deformations occur at the crest level.
This is combined with simultaneous sliding at the concrete to rock interface. Results do indicate that
assumptions made in the non-linear analysis and damping of this dam affect the response. Results are
summarized in Figure 4. Note that stresses within the body of dam change significantly for both cases
involving anchors and no anchors. Instantaneous tension cracks may develop, however tension cracks
will close due to reversal of cyclical earthquake loading. This can further be remedied, if not eliminated,
by installation of post tensioned high strength anchors where a compressive preload along the height of
dam would resist opening of tension cracks. Results indicate that sliding at the concrete-rock interface is
of significant importance as this interface is subjected to maximum static and dynamic loads. Post
tensioned high strength anchors are commonly installed to minimize this upstream-downstream slip and
thus stabilize and increase the safety of the dam during an earthquake.
As both traditional stability and finite element analyses reveal that stabilizing measures are required, a
finite element analysis of the Pine Flat Dam monolith with a vertical anchor is conducted. An anchor
force of 9000 kN is used since this force was estimated from the traditional stability analysis to be the
minimum preload required for stability of the monolith. Thus, a vertical anchor is placed close to the
upstream face of the monolith and the anchor force is distributed uniformly along the height of the
monolith. The presence of high strength anchor across the concrete-rock interface provides a restraint
against translation in horizontal and vertical directions.
As shown in Figure 5, results indicate that maximum average displacement at the concrete-rock drops
from 10 mm to 2 mm for the dam monolith without and with vertical anchors. Energy plots show that
inclusion of anchors dramatically reduces link hysteretic energy absorption at the concrete-rock interface
in friction elements in the upstream downstream direction. The total input energy, which is work done on
the dam monolith by force and earthquake acceleration is also reduced. This is expected as the dam
monolith with the anchor reinforcement is stiffer than the dam monolith without any anchor
reinforcement. Crest displacement is reduced from a peak value of 61 mm to 47 mm for the dam monolith
with anchor reinforcement. Change in average base shear is relatively minimal as this value is influenced
by equilibrium of forces at the concrete-rock interface.
The results of the above finite element analyses will be used in reliability analysis of the Pine Flat Dam
Monolith.

Crest Displacement
(mm)

80

max=61 mm

60
40
20
0
-20
-40
0

10

20

30

40

50

60

50

60

50

60

20

Pmax=-5.347x10 N

Base Shear (x10

N)

TIME (sec)

0
-20
-40
-60
0

10

20

30

40

Base Sliding (mm)

TIME (sec)

12
10
8
6
4
2
0

max=10 mm

10

20

30

40

TIME (sec)

Energy (x 10 6 J)

5000

Total Input Energy

4000
3000

Modal Damping Energy

Energy absorption by friction isolators

2000
1000
0
-1000
0

10

20

30

40

50

60

TIM E (s e c)

Figure 4 Time History Response of Pine Flat Dam without Anchors

Crest Displacement
(mm)

max=47 mm

60
40
20
0
-20
-40
0

10

20

30

40

50

60

N)

TIME (sec)

Base Shear (x10

20
0
-20

Pmax=-5.07x107 N

-40
-60
0

10

20

30

40

50

60

Base Sliding (mm)

TIME (sec)

max=2 mm

3
2
1
0
-1
-2

10

20

30

40

50

60

Energy (x10 6 J)

TIME (sec)

Total Input Energy

5000
4000
3000
2000
1000
0
-1000

Modal Damping Energy

Energy absorption by friction isolators

10

20

30

40

50

60

TIME (sec)

Figure 5 Time History Response of Pine Flat Dam with a vertical Anchor

Reliability Analysis
The stochastic earthquake loading is simplified by considering maximum and critical response values
obtained from the traditional factor of safety and finite element analyses of the Pine Flat Dam Monolith.
A failure function is formulated by considering the free body diagram of forces acting on the dam
monolith with and without anchors. A model showing the primary forces acting on the dam monolith with
a vertical anchor is shown in Figure 6.

C
LAnchor
Reservoir

Panchor

Concrete Dam Monolith

0.78

0.05

mah

Plateral

mg

Pfriction

Idealized Concrete to
Rock Interface

Pfriction
Rock Foundation
Anchor Embedment
into Rock

Puplif t

Rock Cone Mobilized

by Anchor

Figure 6 Primary Forces Acting on The Pine Flat Dam Monolith with Anchor.
For the case involving the determination of factor of safety against sliding by the traditional approach, the
linear failure function is written as:

[11]
where

G = Capacity Demand = vertical

vertical = mg + P
horizontal = P

water

hydrostatic

horizontal

Puplift + Panchor ;
+ Phydrodynamic + Pearthquake ,

and = tan( ) of 1.0 assumed for this example. To consider variability in structural materials and loads,
in causing horizontal and vertical loads and for use in the Relan computer program, relevant mean and
standard deviations are assumed and presented in Table 2.

Table 2 Load and material mean and standard deviation values

Variability

Mean, X i

Standard Deviation, X i

Coefficient of friction,

Moderate

tan() = 1.0

0.10

Concrete material,

Moderate

23.6 kN/m3

2.36 kN/m3

Low

1862 N/mm2

46.55 N/mm2

Item

Anchor strand, fpu

Note: Unit weight of water, w ,of 9.81 kN/ m3 is used as a deterministic variable.

Equation 11 is solved by the Relan computer program for the dam monolith without and with the vertical
anchor and safety indices are listed in Table 3. The term Panchor is zero in equation [11] for the dam
monolith without anchor.
For the case involving the finite element analysis of the dam monolith the failure function is written as:

G = Capacity Demand = vertical Pfriction

[12]

where P friction is the peak base shear response obtained from the non linear finite element analysis of the
Pine Flat Dam monolith. As in the previous case, Equation 12 is solved by the Relan computer program
for the dam monolith without and with the vertical anchor and safety indices are listed in Table 3. For
illustration only, results for time steps yielding maximum base shear with corresponding vertical load and
minimum vertical load with corresponding base shear are shown as a range of safety.
Table 3 Summary of factors of safety and safety indices
Method
Limit
EquilibriumStatic and RELAN
Limit
EquilibriumSeismic and RELAN
Limit
EquilibriumSeismic and RELAN
Finite Element and
RELAN-Seismic
Finite Element and
RELAN-Seismic

Anchor
Provided

Anchor
Load [kN]

Factor of
Safety

Safety
Index

Probability
of Failure

Range of
Safety

No

1.80

3.42

0.416E-03

Survival

No

0.80

-1.874

0.969E-00

Failure

Yes

9,000

1.0

0.00

0.500

Limit States

No

Yes

9,000

-0.651
3.569
1.354
4.804

0.742
0.179E-03
0.878E-01
0.778E-06

Failure
Survival
Survival
Survival

Design Code Approach

For reference only, the National Building Code of Canada specifies that characteristics and probability of
occurrence of the design seismic ground motion are required for seismic design of structures. In this
respect, the probability of exceedance of the seismic ground motion parameters is noted as 10 percent in
50 years, or 0.0021 per annum (1 in 475 years). This is often used by dam safety practitioners in

estimating acceleration records and values for the Design Basis Earthquake (DBE) case. A more rigorous
prbability of exceedance is applied to estimate acceleration records for the Maximum Credible Earthquake
(MCE) case. The application of DBE and MCE cases are not included in the Pine Flat Dam example,
although these characteristics may serve as a reference for development of limit states or reliability based
procedures for analysis and design of concrete gravity or arch dams for that matter.
CONCLUSIONS AND RECOMMENDATIONS
An applied reliability based model in estimating the safety of concrete gravity dam monoliths is outlined
in this paper. As an illustration, the sliding stability of a typical monolith of the Pine Flat Dam is assessed
by the traditional factor of safety method and then compared against results obtained from a two
dimensional finite element and reliability analyses.
It is determined that vertical and horizontal loads calculated in accordance with the traditional limit
equilibrium method are conservative and thus anchor loads calculated are conservative. Although the
results are influenced by assumptions made in the model and modelling of the dam-foundation and damreservoir interactions, results obtained from nonlinear time history analysis of a two dimensional finite
element model illustrate this conservatism.
A comparison of sliding stability analysis results reveals that the traditional sliding factor of safety
approach produces a factor of safety value that is about 25 % below the acceptable limit, whereas the time
history dependant reliability based design approach produces a safety index that, at its lowest time step, is
about 10% below the limit. In addition, anchor load provided to ensure a sliding factor of safety of 1.0
indicates safety at a limit states but a sliding factor of safety of 1.0 does not necessarily indicate safety of
the dam. A small change in analysis or design assumptions may yield factor of safety less than 1.0, thus
revealing ambiguous safety levels.
Assuming load, resistance and importance factors of 1.0 for the earthquake load case and for comparison
purposes, the same anchor load used in the reliability based analysis yields safety of the structure.
Depending on the method used in the stability analysis, this comparison reveals that relatively lower
anchor loads and smaller anchor sizes may potentially be calculated. This ultimately influences the
economics of dam rehabilitation and stabilization.
The use of reliability based design and finite element analysis is recommended for safety evaluation and
design of concrete gravity dams. The same method can be applied in stability analysis and design of arch
dams. This reliability based design also unifies the analysis and design approach employed in reinforced
concrete and steel structures in hydroelectric dam facilities. In parallel with other design codes, it is
further recommended that the above procedure be implemented in dam safety guidelines and design codes
for use by practitioners.
ACKNOWLEDGEMENTS
The author gratefully acknowledges the advice and use of Relan Program developed by Professor Ricardo
Foschi of Department of Civil Engineering, University of British Columbia.
REFERENCES
1. CDA. (1999). Dam Safety Guidelines. Canadian Dam Association, Edmonton, Alberta, Canada.
2. DSI (1998). DYWIDAG Rock and Soil Anchors. DYWIDAG-SYSTEMS International, British
Columbia, Canada.

3. Chavez, Juan W., Fenves, Gregory L. (1994). EAGD-SLIDE: A Computer Program for the
Earthquake Analysis of Concrete Gravity Dams Including Base Sliding. Report No. UCB/SEMM1994/02, Department of Civil Engineering, University of California, Berkeley, California.
4. FERC (1992). Engineering Guidelines for Evaluation of Hydropower Projects. Federal Energy
Regulatory Commission, Washington, D.C.
5. Foschi, R.O., Folz, B., Yao, F. and Li, H. (1997). RELAN: Reliability Analysis Program Users
Manual. Department of Civil Engineering, University of British Columbia, Vancouver.
6. Hansen, K.D., Reinhardt, W.G., Roller Compacted Concrete Dams, McGraw-Hill, New York, 1986.
7. Horyna, T.(1999). Reliability Analysis of Base Sliding of Concrete Gravity Dams Subject to
Earthquakes. Ph. D thesis, Department of Civil Engineering, University of British Columbia,
Vancouver.
8. Kazemi, M. R. (1994). Reliability Based Design of Light Gauge Timber Connectors. M.A.Sc.
thesis, Department of Civil Engineering, University of British Columbia, Vancouver.
9. Li, H., Foschi, R.O. (1998). An Inverse Reliability Method and Its Application. Structural Safety,
Elsevier, 20(3), 257-270.
10. National Building Code of Canada. Supplement to the National Building Code of Canada. Ottawa,
1985.
11. USACE. (1995). Gravity Dam Design. Engineer Manual EM 1110-2-2200, U.S. Army Corps of
Engineers, Washington, D.C., U.S.A.
12. USACE. (1995). Planning and Design of Hydroelectric Power Plant Structures. Engineer Manual
EM 1110-2-3001, U.S. Army Corps of Engineers, Washington, D.C., U.S.A.
13. USACE. (2003). Time-History Dynamic Analysis of Concrete Hydraulic Structures. Engineer
Manual EM 1110-2-6051, U.S. Army Corps of Engineers, Washington, D.C., U.S.A.